Design of process synthesis experiments with a novel distributed algorithm using grids and knowledge-based optimization
Du, D., Cecelja, F., Kokossis, A.
AIChE Annual Meeting, Conference Proceedings
Stochastic optimization algorithms have been developed for decades to offer ways to solve process synthesis problems. However complicated high-throughput applications, which require large amount of components and parameters and serial procedures in many steps, may encounter difficulties with current optimization algorithms, such as slow convergence or being trapped into local optimum. Distributed optimization algorithms, such as the simulated annealing (SA) Cascade, have been invented and introduced for solving large scale process synthesis problems through breaking down long sequential optimization process into smaller and parallel sections by applying grid computing. On the other hand, large amount of idle computing resources, which are widely distributed throughout the world, makes it possible to enable integrated applications and the design of distributed experiments in industrial environments. The paper aims at the design of an efficient high-throughput synthesis infrastructure for seeking higher performance of reactor network design applications with the assist of grids technologies and an attempt to build a self-supervised optimization system on the basis of knowledge-based optimization.